Definitionپێناسە
The work of making AI systems pursue the goals their makers and users intend and follow human values and rules, and not harmful goals of their own. Training with human feedback () is one alignment method.
ئەو کارەی کە سیستەمەکانی وا لێدەکات بەدوای ئەو ئامانجانەدا بچن کە دروستکەران و بەکارهێنەرانیان مەبەستیانە و پەیڕەوی بەها و یاساکانی مرۆڤ بکەن، نەک ئامانجی زیانبەخشی خۆیان. ڕاهێنان بە هەڵسەنگاندنی مرۆڤ () یەکێکە لە ڕێگەکانی.
Exampleنموونە
An aligned assistant refuses to help with a cyber attack even when the request is polite.
یاریدەدەرێک کە یارمەتی هێرشی سایبەری نادات تەنانەت ئەگەر داواکارییەکە بە ڕێزیشەوە بێت.
Other Kurdish renderings in useوەرگێڕانە کوردییەکانی تر
هاوئاڕاستەکردنی ژیریی دەستکردی
Note on the Kurdish termتێبینی لەبارەی زاراوە کوردییەکە
Proposed: هاوئاڕاستەکردن ("making something point the same way", ئاڕاستە as in the registry's ئاڕاستەبڕ), like Persian همسوسازی.